This story actually starts a few weeks back in Davos, at the World Economic Forum. On one of the panels there, some global voices, including the IMF, hinted that India is still “second tier” in AI, far behind the US and China. Ashwini Vaishnaw, India’s IT and railways minister, pushed back strongly on that idea.
He said that India is not behind, and cited Stanford rankings that place India among the top three countries in AI preparedness and penetration, and number two in terms of AI talent. His core argument was simple: Don’t look only at who is spending the most money; look at who is building a full AI stack and real-world solutions.
Vaishnaw also used Davos to spell out what this AI strategy actually looks like. According to him, India wants “sovereign AI” – meaning the country should not be dependent on any one foreign company or country for core AI capabilities.
He explained that India is working across all five layers of the AI stack: applications that people and companies use, models that power those applications, chips, physical infrastructure like data centres, and energy to run all of it. The pitch was clear: India is not trying to copy the US or China model; it is building its own path, focused on returns, inclusion and real impact.
Cut to Delhi now, where this narrative is being tested on home ground at the India–AI Impact Summit 2026 at Bharat Mandapam. The summit is bringing together leaders from over 100 countries, global CEOs, and tens of thousands of delegates to discuss how AI can be used for development, governance, and business.
It is the first large, truly global AI summit hosted in the Global South, and the government wants to use it to show that India is ready to sit at the “AI high table” alongside the US and China, not somewhere in the background. The expectation is that big-ticket investment announcements will also follow, especially around data centres, cloud and AI applications, signalling that global capital is taking India’s AI story seriously.
One big reason India can make this claim today is that its homegrown AI talent and models are starting to show results. A recent Karnataka-linked or India-focused benchmarking exercise and media reports have highlighted how Indian models like Sarvam AI are beating or matching global giants on some Indian use cases. In a key OCR benchmark that tests how well AI reads real-world documents, Sarvam Vision scored around 84 per cent accuracy, ahead of OpenAI’s ChatGPT-based systems and Google’s Gemini 3 Pro.
On another benchmark for complex document layouts and tables, Sarvam again crossed 93 per cent, showing that a local model trained on Indian conditions can actually do better than global models for specific tasks. This gives the government a strong example to show that “sovereign AI” is not just a slogan; there are real products emerging.
But India’s AI story is more complicated than just who has the best model. The country has over 20 major languages and hundreds of dialects, plus very different levels of income, education and digital access. That means any AI that works well only in English, or only for urban elites, is not enough. This is why there is a big push on multilingual AI – models and apps that can understand and respond in Indian languages, and also preserve local culture and knowledge.
Tools like Sarvam’s Bulbul voice model, which supports voices across many Indian languages, are an example of how AI can help keep languages alive instead of pushing everyone towards one global language. For India, language is not just a technical issue; it is an emotional and cultural one, so the AI strategy has to respect that.
The core objective of the India AI Impact Summit fits into this larger picture. The idea is not only to talk among Indians, but to plug India’s AI ecosystem into the global value chain. That means three things: showing the world what Indian startups, researchers and companies are building; attracting more investment, partnerships and compute to India; and shaping global rules on responsible and inclusive AI from a Global South perspective.
The summit is being framed around “People, Planet and Progress” – which basically means AI that helps citizens, respects the environment and supports fair growth instead of just making a few companies richer.
Prime Minister Narendra Modi’s role in this story is also central. For almost a decade, he has tried to brand India as a digital power through Aadhaar, UPI, DigiLocker and other platforms, and many tech CEOs have publicly praised his grasp of technology and AI policy.
At Davos-type forums and bilateral meetings, global AI leaders have often said that Modi understands the long-term implications of AI – from data governance to jobs – better than many other world leaders. At the Delhi summit, he is expected to use that image to position India as a responsible AI leader – a country that can combine scale, democratic values and digital public infrastructure to offer an alternative to both the US Big Tech-dominated model and the Chinese state-led model.
On the ground, both the government and private sector are working to back this narrative with real moves. The government has launched the IndiaAI Mission and is funding a set of sovereign AI models with tens of billions of parameters focused on priority sectors like agriculture, health, governance and climate.
It has set aside money for compute infrastructure, design-linked incentives for chip and GPU design, and is working with players like Nvidia to accelerate domestic capability. At the same time, global giants such as Google, Amazon and Microsoft have committed large investments to expand data centres and cloud capacity in India, because they see it as one of the fastest-growing AI markets in the world.
The private ecosystem is also warming up. Venture capital funds focused on Indian AI are raising more money, and India’s IT services industry, which employs millions, has started to pivot from pure labour-based outsourcing to AI-powered services and products.
Indian startups are experimenting with everything from AI for documents and voice to AI for healthcare, agriculture and governance. Still, the big open question is whether these startups can scale fast enough, reach meaningful revenue, and convince Indian companies and government departments to buy from them instead of sticking to foreign vendors.
Finally, there is the uncomfortable but important job question. India’s tech sector employs millions of people, and AI will automate some of the work they do today. Government think tanks have already warned that if India does nothing, a significant chunk of these jobs could be at risk in the next few years. The official line is that with the right skills, reskilling and new AI-related roles, India can absorb this shock and even create new kinds of work.
But managing this transition – making sure ordinary workers are not left behind while India chases its “seat at the AI high table” – may turn out to be the toughest part of the entire AI revolution for India.
